China treats a machine learning algorithm tied to hardware performance or to natural-law data as technical

Patent Examination Guidelines (2023), China National Intellectual Property Administration Order No. 78, issued 21 December 2023, effective 20 January 2024, Part II, Chapter 9, Section 6

The China National Intellectual Property Administration revised the chapter of its Patent Examination Guidelines on inventions involving computer programs to give clearer rules for big data and artificial intelligence. Under the revised subject-matter test, a solution that improves an artificial intelligence or big data algorithm, such as deep learning or clustering, is a technical solution under Article 2(2) of the Patent Law where the algorithm has a specific technical association with the internal structure of the computer system and solves the technical problem of improving hardware computing efficiency or execution effect, for example by reducing data storage or transmission or increasing processing speed. Big data processing in a specific application field also qualifies where classification, clustering, regression analysis or neural networks are used to mine internal relationships in the data that conform to natural laws. At inventive step, the contribution of an algorithm feature that functionally supports and interacts with the technical features must be considered, and an improvement in user experience counts only if it is produced by the technical features, alone or together with such interacting algorithm features. New examples address training a deep neural network, coupon usage propensity, knowledge-graph reasoning and adapting neural network parameters, and hold a financial product price prediction method not to be a technical solution.

More on Eligibility of software and AI claims